Abstract

Keywords: veterinary cordon fence | ungulate | fence ecology | movement | fence interaction | GPS

Introduction

In most semi-natural landscapes wildlife is unavoidably confronted with fences (Jakes et al. 2018). Interactions between animals and wildlife-proof fences are very common and increase the risk of entanglement and mortality in wildlife (e.g., Mbaiwa and Mbaiwa 2006; Safner et al. 2021). Especially in southern Africa such fences are prevalent since wildlife-based land-use is a common and growing practice (e.g., P. A. Lindsey, Romañach, and Davies-Mostert 2009; Taylor et al. 2020), conservation success often depends on them (e.g., Hayward 2012) and Veterinary Cordon Fences (VCFs), established to minimize wildlife-livestock disease transmission (e.g., foot-and-mouth disease, Caron et al. 2013), spread over thousands of kilometers in Zimbabwe, Botswana and Namibia (Cumming et al. 2015). Despite fatal effects of fences (e.g., Spinage (1992)), most interactions between larger mammals and fences lead to altered behavior such as overexploitation of foraging resources You et al. (2013), narrowed stop-over areas during seasonal migrations (e.g., Sawyer et al. 2013), or partly increased hunting activities in closer proximity to fences (Dupuis-Desormeaux et al. 2016). Wildlife-proof fences are an effective measure to suppress movements of large, non-flying vertebrates, but the urge to cross remains, especially in migratory species, and gaps often occur. Despite human-made damages or gaps (Dupuis-Desormeaux et al. 2018), selective passages (e.g., tubular metal grills (Gates et al. 2012), rock walls (Dupuis-Désormeaux et al. 2016)), or selective fence designs (e.g., Weise et al. 2011; Laskin et al. 2020), passages through fences are created by animals either by undermining (Kesch, Bauer, and Loveridge 2014) or by directly damaging the fence. Mainly elephants cause the latter (e.g., Mutinda et al. 2014; Mogotsi, Kgosikoma, and Lubinda 2016) whereby they open the barrier for all other species, which heralds the role of megaherbivores as ecosystem engineers in the Homogenocene.

Animal-fence and fence-gap interactions are commonplace in the semi-natural landscapes of subtropical Africa (Naidoo et al. 2022), however our knowledge about the direct effects of such on the animals’ movement and behavior patterns is still limited. Clearly, the majority of animal species interacts with fences once these are located in their homeranges (e.g., Kesch, Bauer, and Loveridge 2014; Wilkinson et al. 2021), but larger (> 1 kg) herbivores are affected to a large degree since many species travel long distances to reach forage and water resources (e.g., Bartlam-Brooks, Bonyongo, and Harris 2011; Naidoo et al. 2022), especially under semi-arid conditions with often scattered and erratic resources (e.g., Norman Owen-Smith 2014; Ibrahim et al. 2021). By classifying animal behavior based on camera trap data at Lake Nakuru National Park, Kenya, Wilkinson et al. (2021) show that ungulate - fence interactions were most frequently observed, especially during the day. They identified major behaviors at fences to be cross, straddle (individuals on opposite fence side), grazing/drinking and vigilance. Xu et al. (2021) quantified ungulates’ behavior at fences based on hourly GPS-telemetry data of two species in the rangelands of Wyoming, USA and found mainly quick interactions (cross or bounce) to occur, but also average movement, tracing, back-and-forth, and trapped - like behaviors were identified. Further studies examine crossing behaviors (e.g., Palmer et al. 1985; Harrington and Conover 2006; Burkholder et al. 2018; Segar and Keane 2020), but studies examining alteration of ungulate movements directly at fences are scarce (e.g., Goddard et al. 2001; Gulsby et al. 2011; Nandintsetseg et al. 2019), especially with regard to exact measures of locomotion.

In the light of one of todays’ major goals in nature conservation, namely the preservation and restoration of ungulate migrations with connected substantial ecosystem processes (e.g. Kauffman et al. 2021a), the improvement of our knowledge on the direct effects of fences on ungulate movement is essential. Fence and fence-gap planning and designing as well as policy making require a fundamental understanding of how ungulates interact with fences, to which direct effects on locomotion this leads and what ultimate consequences emerge, especially with regard to the energetic losses the animals are confronted with. One step to disentangle the complexity of fence effects on wildlife is to evaluate altered locomotion of ungulates at fences and fence-gaps.

In this study we are using a natural experimental set-up with three GPS-collared southern African ungulate species, differing in movement strategies (range residency: greater kudu - Tragelaphus strepsiceros, partial migration: springbok - Antidorcas marsupialis, nomadism: common eland - Tragelaphus oryx) and living in a semi-arid savanna landscape where privately owned land borders Etosha National Park. The areas are separated by Namibia’s VCF and other non-electrified and electrified wildlife-proof fences. The GPS-collars were put on 25 individuals and recorded localizations for over two and half years every 5 to 15 minutes so that single fence interactions can be identified and analyzed with high precision. The VCF is regularly being damaged by elephants and other wildlife and created gaps are frequently used by the animals to move between the National Park and the privately owned land. We use a twofold approach to (I) disentangle types and times of animal-fence interaction, and based on this (II) to investigate direct effects of wildlife-proof fences on the antelopes’ locomotion. For the latter we relate times and distances spend at fences to individual movement strategies, and evaluate direct effects on locomotion (indicated by speed) with regard to the types of animal-fence interaction.

Methods

Study area

The study area is located 80 km south-west of the Etosha pan (15.2235°E, 19.2576°S, Fig.: 1) at the boundary of the regions Kunene, Omusati and Oshana in Namibia’s North. The core area is a private wildlife reserve (‘Etosha Heights’ - EH, 460 km²) which adjoins the South of Etosha National Park (ENP, 22941 km²). These two areas are separated by a 70 km long section of Namibia’s VCF. EH adjoins private farms in the West, South and East separated by a well maintained, electrified, and wildlife-proof fence line (2.4 m high, 18-strand). On two of these farms further movement data was captured. The farms’ outer borders consist also of non-electrified wildlife-proof fences (2.4 m high, multi-strand, lower half with wire mesh).

Study area with fence lines (black line - veterinary cordon fence, orange line - electrified fence, blue lines - wildlife-proof fences), movement trajectories (in yellow) and sub-trajectories 45 minutes before, during and after fence interactions (colored lines). Each individuals' position, which is farest away from its most visited water point (blue points) is shown by species icons. Background Sentinel 2 (Bands 3,4,5) image, March 2020 (Contains modified Copernicus Sentinel data [2020]). Upper left Namibia map shows veterinary cordon fence (VCF, red line) and focal wildlife conservation areas (yellow area - Etosha National Park, red point - Etosha Heights)

Figure 1: Study area with fence lines (black line - veterinary cordon fence, orange line - electrified fence, blue lines - wildlife-proof fences), movement trajectories (in yellow) and sub-trajectories 45 minutes before, during and after fence interactions (colored lines). Each individuals’ position, which is farest away from its most visited water point (blue points) is shown by species icons. Background Sentinel 2 (Bands 3,4,5) image, March 2020 (Contains modified Copernicus Sentinel data [2020]). Upper left Namibia map shows veterinary cordon fence (VCF, red line) and focal wildlife conservation areas (yellow area - Etosha National Park, red point - Etosha Heights)

The climate of the study region is semi-arid with highly variable rainfall mainly occurring from September to April (green season; mean temperature: 26°C). The colder dry season lasts from May to August (mean temperature: 18°C). In the study area mean annual precipitation increases from south-west to north-east and is between 250 to 350 mm (averaged from 1981-2017 with CHIRPS data, Funk et al. (2015)).

Vegetation in the study area is heterogeneous and ranges from mountainous mixed woodlands (dominated by Terminalia prunioides, Commiphora glandulosa and Combretum apiculatum) over to Colophospermum mopane wood- and shrublands, Catophractes alexandrii and Vachellia nebrownii shrublands up to grasslands dominated by Enneapogon desvauxii, Aristida adscensionsis and Eragrostis nindensis. During dry season ground is mainly bare. While woody species foliate more periodically (Archibald and Scholes 2007; Ibrahim et al. 2021) with slight differences caused by local hydrological conditions, green-up of herbaceous plants is dominantly triggered by local rainfall events and thus occurs erratic and patchy.

Study species

We were particularly interested in differences between large herbivore species of diverse movement types (Bunnefeld et al. 2011; Kauffman et al. 2021b) and focused on females since they form larger herds. All choosen species are common in the study region and of economical importance (Peter A. Lindsey 2011). Eland (Tragelaphus oryx) have largely been re-established on private reserves. Kudu (Tragelaphus strepsiceros) and springbok (Antidorcas marsupialis) mostly occur naturally on wildlife, mixed and even cattle farms. On Etosha Heights only eland were relocated from the Waterberg Plateau Nationalpark (250 km south-east of the study area) in the early 2000s. While springbok primarily crawl under fences and, consequently, are already impeded by cattle fences with closely spaced strands or wire mesh (Bigalke 1972), eland and kudu jump over fences up to a height of 2 m (N. Owen-Smith 1985) and more. The species are non- or low-water reliant (Hempson, Archibald, and Bond 2015), but during dry season they use water holes frequently, if available.

We selected the springbok as medium-sized (weight of female: 37 kg (Skinner and Chimimba 2005)), non-water reliant mixed feeder (Hempson, Archibald, and Bond 2015) living in small groups during dry season and aggregating in large groups during the green season (Skinner and Chimimba 2005). Migratory movements of springbok vary largely across years and individuals. For instance, the historical huge “treks” of hundreds of thousands of springbok, observed in the 19th and 20th centuries (e.g., cronwright1931?) in southern Africa, are attributed to an interplay between successive years of high rainfall, high population growth, drought and overgrazing, causing mass movements (e.g., Roche 2005). The movement patterns of the collared individuals varied from range residency to migration (Fig.: 2).

Overview of the collared individuals' diverse movement patterns. Within species-wise panels (from left to right: eland, kudu, springbok) displacement is shown on y-axis, measured as distance in km to most visited waterpoint, along the observation time (x-axis, note different overall observation times). Individual patterns are marked by different colors.

Figure 2: Overview of the collared individuals’ diverse movement patterns. Within species-wise panels (from left to right: eland, kudu, springbok) displacement is shown on y-axis, measured as distance in km to most visited waterpoint, along the observation time (x-axis, note different overall observation times). Individual patterns are marked by different colors.

We further choose the nomadic eland, which is the largest African antelope (weight of females: 305 kg). Eland is classified as social, non-water reliant mixed feeder (Hempson, Archibald, and Bond 2015) and the group sizes vary as those of springboks. They are mainly nomadic, but range residency is known as well as migratory movements (Hillman 1988). The collared individuals’ movement patterns were mainly full nomadic with partly irruptive sequences of long distance movements (Fig.: 2).

The greater kudu is a large (weight of females: 152 kg (Skinner and Chimimba 2005)), social, low water dependent browser living in small, relatively constant groups of less than 14 females (Skinner and Chimimba 2005). Kudu is sedentary with rare emigrations to other groups [@ Owen-Smith1990]. Except for one, all collared kudus’ movement patterns were range resident (Fig.: 2).

Tracking data

Individuals were collared with e-Obs GPS devices (e-Obs GmbH, Germany; springbok: Collar 1D with 320 g, eland and kudu Collar Big 3D with 840 g or Collar Big 4D with 960 g). Each individual was chosen from a different dry season group to avoid spatio-temporal overlap of the tracking data and cover the movement of many individuals per tracking device. Animals were chemically restrained from ground (springbok) or from a helicopter (kudu and eland). Details on the capturing procedure can be found in (Hering2022?). Animal handling permits were approved by the Namibian National Commission On Research Science & Technology, certificate number RCIV00032018 (authorisation numbers: 20190602, AN20190808).

GPS localizations were taken in bursts at 15, 7.5 or 5 minute intervals depending on battery size of the respective collar (Tab.: A-??). A burst consisted of at least three consecutive GPS localizations (one second interval). These GPS bursts were averaged for the analysis.

Identification and definition of fence interaction

All analyses were performed with R (version 4.2.0, (R-base?)). We used the same procedure of fence interaction identification as described in (Hering2022?) in detail. In a first step fence lines were buffered by 50 m yielding areas of 100 m width to account for potential inaccuracy of the geo-data. In a second step, tracks of fence interaction were identified by selecting subsequent GPS-localizations, which overlapped with the buffered fence areas. We allowed consecutive positions with a time threshold of 60 min to be outside the fence areas if the former and next were inside, assigning them to one single fence interaction event. Events with less than two localizations were classified as quick and all others as “long” (example see Fig.: 3).

Types of fence interaction were adapted from Xu et al. (2021). We distinguished basic interactions into cross or non-cross events. Basic interactions were classified depending on movement behavior and time spent at a fence into four interaction types: (i) quick, as immediate interaction, (ii) trace (no changes in direction), (iii) back-and-forth (changes in direction, at least one relative angle larger than 120°), and (iv) stay (at minimum one event of consecutive speeds below 0.5 m/s lasting longer than 15 minutes , independent of directional change). Quick fence interactions events consisted either out of the two localizations with the fence area between (no localization was at the fence area) or out of the one position at the fence area. The interaction type stay was added to the original types introduced in Xu et al. (2021).

Example of the identification of fence interaction types. Shown are examples for each of the classified types of fence interaction (see panels titles, species of sample track indicated by icon). The event itself is marked in green. Arrows show the direction of movement paths. Grey areas are the fence area (fence line with a 50 m buffer).

Figure 3: Example of the identification of fence interaction types. Shown are examples for each of the classified types of fence interaction (see panels titles, species of sample track indicated by icon). The event itself is marked in green. Arrows show the direction of movement paths. Grey areas are the fence area (fence line with a 50 m buffer).

Direct effects of wildlife-fence interactions

Results

Abundances of fence interaction types

We identified a total of 3582 single fence interaction events out of which 72 % were observed at the veterinary cordon fence (VCF), 21.5 % at the electrified fence (EF) and 6.5 % at the non-electrified, wildlife-proof fences (NEF). Nine out of ten collared springbok encountered the VCF, seven the EF and one a NEF. All of the seven eland encountered the VCF as well as the EF, while for five of them interactions with a NEF were identified. We found three out of the eight kudu which encountered the VCF, six encountered the EF and two were found at a NEF. Lowest number of total encounters was recorded for a kudu (id: 7296) with six encounters during 559 days of observation. Highest number of total encounters was recorded for another kudu (id: 7295) with 436 encounters during 526 days of observation. Average daily encounters ranged from 0.3 ± 0.2 SD for springbok, over 0.3 ± 0.3 SD for kudu up to 0.5 ± 0.2 SD for eland.

Total numbers of fence encounters grouped row-wise by fence type (upper row: **veterinary ** cordon fence; lower row: **electrified** fence. Number of encounters (x-axis) shown per fence interaction type (y-axis) with individual numbers indicated by color. Outer lines of bars indicate crossing success (black - cross; grey - non-cross). Species-wise numbers shown per column (see species-symbol).

Figure 4: Total numbers of fence encounters grouped row-wise by fence type (upper row: veterinary cordon fence; lower row: electrified fence. Number of encounters (x-axis) shown per fence interaction type (y-axis) with individual numbers indicated by color. Outer lines of bars indicate crossing success (black - cross; grey - non-cross). Species-wise numbers shown per column (see species-symbol).

Major identified fence interaction types of sub-type cross were quick cross at the VCF for eland (26 % of encounters by eland) and springbok (29 % of encounters by springbok), and trace cross at the VCF for kudu (12 % of encounters by kudu). Major interaction types of the sub-type non-cross were stay non-cross at the VCF for eland (12 % of encounters by eland), quick non-cross at the VCF for springbok (15 % of encounters by springbok) and stay non-cross at the EF for kudu (23% of all encounters by kudu). Further frequent fence interactions were of type trace for all species (Fig.: 4).

Times of fence interactions

We found peaks in number of fence interactions for all species during the transition from dry to green season (Fig.: 4). During this time average above ground woody biomass of visited locations increased for all species, indicating a temporal shift in forage resources. Lowest number of fence interaction for eland and kudu were recorded during the transition from green to dry season and for springbok during the middle of the dry season.

Temporal peaks in numbers of fence interactions along seasons and in relation to environmental conditions. **Upper row of panels** shows species-wise (panels, see icon) and interaction type specific (see legend) counts of identified interactions per season (60-day periods). **Lower row of panels** shows daily precipitation (blue lines) and daily average of above ground woody biomass of the locations which were visited by all individuals of a species (brown line). Low values of woody biomass indicate open habitats, high values indicate shrub - and woodlands. Note that for the last season (mid green 21) tracking data was available for a few individuals only.

Figure 5: Temporal peaks in numbers of fence interactions along seasons and in relation to environmental conditions. Upper row of panels shows species-wise (panels, see icon) and interaction type specific (see legend) counts of identified interactions per season (60-day periods). Lower row of panels shows daily precipitation (blue lines) and daily average of above ground woody biomass of the locations which were visited by all individuals of a species (brown line). Low values of woody biomass indicate open habitats, high values indicate shrub - and woodlands. Note that for the last season (mid green 21) tracking data was available for a few individuals only.

The majority of all fence interactions was detected during the day (from start of dawn to end of dusk). This was for eland 69% of all interactions (52% only at the VCF), for kudu 75% of all encounters (46% only at the VCF, mainly one individual), and for springbok 92% of all encounters (83% only at the VCF). Fence interactions at the VCF occurred mainly during morning times (Fig.: 6). Number of interactions peaked two to two and a half hours after dawn (70 to 100 minutes after start of sunrise) for eland, two and half to three hours (100 to 130 minutes after start of sunrise) for kudu, and three to three and a half hours (130 to 160 minutes after start of sunrise) for springbok. Secondary peaks were pronounced in eland (12 to 12.5 hours after dawn) and springbok (11.5 to 12 hours after dawn) at the VCF and for kudu (12 to 12.5 hours after dawn) at the EF. Temporal patterns of fence interactions at the VCF and during the day mainly corresponded to overall patterns of directed movements during the day (Fig.: 6).

Temporal peaks in numbers of fence interactions along times of the day. Shown are density distributions of times of observed fence interactions (violet lines) and of times of directed movements (grey dotted lines, speeds larger than 0.5 m/s and relative angles less than 90°) per species (columns of panels, see symbol). Density (y-axis) for times after start of dawn (x-axis, sun 12° below horizon). Vertical lines indicate time of sunrise (left line) and times of sunset (earliest sunset indicated by dotted line, latest sunset shown by right line). Note: Statistical population is either species-wise fence encounters (violet density estimates) or all directed GPS-localizations (grey density estimates, 8 % of all kudu localizations, 13 % of all springbok localizations, 18 % of all eland localizations)

Figure 6: Temporal peaks in numbers of fence interactions along times of the day. Shown are density distributions of times of observed fence interactions (violet lines) and of times of directed movements (grey dotted lines, speeds larger than 0.5 m/s and relative angles less than 90°) per species (columns of panels, see symbol). Density (y-axis) for times after start of dawn (x-axis, sun 12° below horizon). Vertical lines indicate time of sunrise (left line) and times of sunset (earliest sunset indicated by dotted line, latest sunset shown by right line). Note: Statistical population is either species-wise fence encounters (violet density estimates) or all directed GPS-localizations (grey density estimates, 8 % of all kudu localizations, 13 % of all springbok localizations, 18 % of all eland localizations)

Fence interaction in relation to use of space

We did not find an indication for a relationship between individuals’ use of space and the time or distance they spend at fence-lines. Species-wise maxima of average weekly proportion of time and distance at fence-lines were identified for individuals with an intermediate space-use area of 232 km² to 290 km² (Fig.: 7). These were kudu 7295 (used area: 286 km², with 5.6 %, 95% CI [4.4, 6.8] out the averagely 41.3 km, 95% CI [38.3, 44.4] travelled per week and 4.5 %, 95% CI [3.4, 5.7] of weekly time spend at fences), springbok 8317 (used area: 232 km², with 3.1 %, 95% CI [1.6, 4.6] out the averagely 71.5 km, 95% CI [61.6, 79.4] travelled per week and 2.3 %, 95% CI [1.2, 3.5] of weekly time spend at fences) and eland 7302 (used area: 290 km², with 5.9 %, 95% CI [3.7, 8.1] out the averagely 87.4 km, 95% CI [79.4, 94.7] travelled per week and 3.7 %, 95% CI [2.6, 5.1] of weekly time spend at fences).

On average the animals travelled between a minimum of 0.2 %, 95% CI [0.01, 0.6] (kudu 7296, used area of 33 km² with all of it inside the fenced-in area) and a maximum of 5.9 %, 95% CI [3.7, 8.1] (eland 7302, used area of 290 km² with 51% of it inside the privately owned areas) of their weekly distance covered along fence lines. Species-wise averages of weekly travelled distances along fence-lines as percentage of total weekly distance covered were 1.9 %, 95% CI [1.6, 2.2] for kudu (573 weeks of observation, 0.8 km / week, 95% [0.6,0.9]), 1.0 %, 95% CI [0.8, 1.1] for springbok (709 weeks of observation, 0.7 km / week, 95% CI [0.6, 0.8]), and 2.9 %, 95% CI [2.5, 3.4] for eland (432 weeks of observation, 2.6 km / week, 95% CI [2.2, 3.0]). Noteworthy, the highest proportion recorded was 32 km of distance travelled along fence lines out of 86 km distance covered (37 %) for an eland (eland 7302) during the first week of August 2020 (Fig.: 8). Longest weekly distance along fence-lines for springbok was 12.8 km out of a total of 102.0 km and 11.9 km out of 43.3 km for a kudu.

The averages of the amount of weekly time spend at fence-lines ranged between a minimum of 0.2 %, 95% CI [0.2,0.7] (kudu 7296) and a maximum of 4.5 %, 95% CI [3.4, 5.7] (kudu 7295). Species-wise averages of time spend at fence-lines as percentage of total time were 1.5 %, 95% CI [1.3, 1.8] for kudu, 0.7 %, 95% CI [0.6, 0.8] for springbok, and 2.6%, 95% CI [2.1, 3.0] for eland. Noteworthy, the longest weekly time spend at fence we recorded was 48.4 hours for an eland (eland 7287) during the first week of February 2021 within the private farm, adjacent to the private wildlife-reserve (Fig.: 8. Longest weekly time at fences for springbok was 20.3 hours (springbok 6769) and for 40.1 hours for kudu (kudu 7295).

Weekly distances (A) and times (B) individuals spend at fence lines projected on the area they used. Shown are mean values (symbol position) on y-axis separated by individual and species (individual indicated by color, species indicated by symbol) with lower (0.025) and upper (0.975) confidence intervals from bootstrapping (1000 iterations) shown by black, vertical lines. Individuals are ordered by the area they used (x-axis, area based on 95% of kernel density estimates with ad hoc method for smoothing parameter calculation; peak dry season localizations excluded). Single points depict values of single weeks of observation. Note, y-axis is limited to 20; data is not. X-axis is squareroot transformed.

Figure 7: Weekly distances (A) and times (B) individuals spend at fence lines projected on the area they used. Shown are mean values (symbol position) on y-axis separated by individual and species (individual indicated by color, species indicated by symbol) with lower (0.025) and upper (0.975) confidence intervals from bootstrapping (1000 iterations) shown by black, vertical lines. Individuals are ordered by the area they used (x-axis, area based on 95% of kernel density estimates with ad hoc method for smoothing parameter calculation; peak dry season localizations excluded). Single points depict values of single weeks of observation. Note, y-axis is limited to 20; data is not. X-axis is squareroot transformed.

Example weekly movement paths of weeks with either very long distances moved along fence lines (left) or times spend at fence lines (right). Red points depict localizations related to a fence-interaction event. Further details are shown in panel titles or within the legends. Background Sentinel 2 (Bands 3,4,5) image, March 2020 (Contains modified Copernicus Sentinel data [2020]).

Figure 8: Example weekly movement paths of weeks with either very long distances moved along fence lines (left) or times spend at fence lines (right). Red points depict localizations related to a fence-interaction event. Further details are shown in panel titles or within the legends. Background Sentinel 2 (Bands 3,4,5) image, March 2020 (Contains modified Copernicus Sentinel data [2020]).

We found correlations between the index of weekly hole fidelity and distances travelled along fence-lines for all species (Fig.: 9. The GLM (on Gamma-distributed response variable with reciprocal link-function) yielded a residual deviance : degrees of freedom ratio of 0.95 and was superior to the intercept model only (adjusted McFadden’s pseudo R²: 0.11). The estimates on the effect of hole fidelity on weekly distance travelled along fence-lines were highest for eland (0.76 ± 0.1 SD) and low for kudu (1.3 ± 0.2 SD) and for springbok (1.3 ± 0.1 SD).

Relation between species-wise weekly distances moved along fence-lines and hole fidelity. Hole fidelity (x-axis) as index of maximum amount a fence section was used by an individual for crossing in relation to all weekly used sections for crossing. High values indicate high hole fidelity. Single points (colors depict single individuals) show weekly distances (y-axis). Right of vertical dotted line: species-wise (in panels, see symbol) prediction from GLM (black lines, Gamma distribution with inverse link function) with standard error (grey areas). Left of vertical dotted line: distances along fence lines of weeks without crossing (not used for prediction).

Figure 9: Relation between species-wise weekly distances moved along fence-lines and hole fidelity. Hole fidelity (x-axis) as index of maximum amount a fence section was used by an individual for crossing in relation to all weekly used sections for crossing. High values indicate high hole fidelity. Single points (colors depict single individuals) show weekly distances (y-axis). Right of vertical dotted line: species-wise (in panels, see symbol) prediction from GLM (black lines, Gamma distribution with inverse link function) with standard error (grey areas). Left of vertical dotted line: distances along fence lines of weeks without crossing (not used for prediction).

Fence effects on motion

We found differences in speed when animals crossed the VCF for most of the identified combinations of species and fence interaction type.

For all species quick cross events (number of total events: kudu - 118, springbok: - 306, eland - 371) resulted in increased speeds while crossing compared to speeds immediately before or after the crossing; whereas speeds before and after did not differ (Fig.: 10). On average kudu were 1.6-times (0.14 m/s, 95% CI[ 0.10, 0.17]), springbok 1.4-times (0.14 m/s, 95% CI [0.10, 0.18]) and eland 1.3-times (0.19 m/s, 95% [0.15, 0.22]) faster during quick cross as compared to average speeds 45 minutes before they crossed.

For kudu also trace cross events resulted in increased speeds when they were close to the fence (131 events, 1.2-times faster; 0.04 m/s, 95% CI [0.02,0.07]). Average speeds of springbok after crossing differed from those before and during the trace cross fence interactions (100 events) and were 1.5-times faster than before (0.15 m/s, 95% CI [0.09, 0.21]). Average speeds of eland differed before and during trace cross events (180 events) as well as before and after the event. On average eland were 1.4-times faster (0.16 m/s, 95% CI [0.10, 0.22]) after the fence interaction as compared to average speeds before.

Average speeds during fence interactions of the type stay cross differed from those before and after the event in springbok (59 events) and eland (91 events) but not in kudu (98 events). On average speeds during the fence interaction were 2.2-times slower (-0.17 m/s, 95% CI [-0.24, -0.12]) than before for springbok and 2.1-times slower (-0.2 m/s, 95% CI [-0.26, -0.14]) than before for eland.

We found more than 50 back-and-forth cross events only for eland. The average speeds did not differ between before, during, or after the event.

For non-cross fence interactions at the VCF with more than 50 observed events we found the following (all results can be found in the supporting material). Average speeds related to quick non-cross differed between during and after the interaction as well as before and after in springbok (158 events, 1.4 times slower after the event than before: -0.04 m/s, 95% CI [-0.07, -0.02]). We did not find differences in average speeds for trace non-cross events in springbok and eland. For stay non-cross events we found differences in average speeds for all species. In kudu average speeds differed before and during the events (138 events) but average differences in speed were low (< 0.01 m/s). Average speeds in springbok (114 events) were 2.3-times slower (-0.11 m/s, 95% CI [-0.14, -0.08]) during the event as compared to before. Average speeds in eland (169 events) were 2.9-times slower (-0.17 m/s, 95% CI [-0.22, -0.14]) during the event as compared to before. We did not find differences in average speeds for back-and-forth non-cross events.

For non-cross fence interactions at the EF with more than 50 observed events we found the following (all results can be found in the supporting material). Average speeds of kudu during trace non-cross events (135 events) differed from those before and after the event. On average speeds were 1.3-times faster (0.04 m/s, 95% CI [0.03, 0.06]) during the event than before. Speeds of kudu during stay non-cross differed from those before (247 events). They were 1.5 times slower (-0.03 m/s, 95% CI [-0.04,-0.02]). Also for eland differences between average speeds before, during and after stay non-cross events (96 events) were identified and speeds during the fence interaction were 2.1-times slower (-0.1 m/s, 95% CI [-0.13,-0.07]) as compared to before.

Only one eland (eland 7287) interacted with a NEF frequently. Dominantly (101 events of a total of 177 events) stay non-cross fence interactions were identified. Thereby average speeds during the interaction were 2.5-times slower (-0.11 m/s, 95% CI [-0.14, -0.08]) as compared to before.

Results for comparisons of speeds at the VCF and related to crossings. Comparison of average speeds (single observations as colored points, color indicates individual) before, during and after a fence interaction (all speeds 45 min before and after the respective fence interaction) and during the interaction (all speeds during the fence interaction) with Box-Whisker-Plots (no outliers shown, confidence region indicated by notches). Speeds on y-axis and temporal groups on x-axis. Single observations are connected by a line. Columns of panels refer to species (see symbol in first row); rows of panels refer to fence interaction type (see row titles on the right). Results of heteroscedastic one-way repeated measures (before, during, after) ANOVA for trimmed means of speeds indicated on horizontal lines (lines depict compared couple of temporal groups) with corresponding significance codes: p > 0.05: n.s. (not significant); p < 0.05: `*`; p < 0.001: `**`; p < 0.0001: `***`

Figure 10: Results for comparisons of speeds at the VCF and related to crossings. Comparison of average speeds (single observations as colored points, color indicates individual) before, during and after a fence interaction (all speeds 45 min before and after the respective fence interaction) and during the interaction (all speeds during the fence interaction) with Box-Whisker-Plots (no outliers shown, confidence region indicated by notches). Speeds on y-axis and temporal groups on x-axis. Single observations are connected by a line. Columns of panels refer to species (see symbol in first row); rows of panels refer to fence interaction type (see row titles on the right). Results of heteroscedastic one-way repeated measures (before, during, after) ANOVA for trimmed means of speeds indicated on horizontal lines (lines depict compared couple of temporal groups) with corresponding significance codes: p > 0.05: n.s. (not significant); p < 0.05: *; p < 0.001: **; p < 0.0001: ***

We found a positive correlation between the distance springbok or eland travelled along the VCF before they crossed and the speeds with which they moved after they crossed (average speeds 45 min after the fence interaction). We found this for situations when they moved out of Etosha National Park (Fig.: 11). When they moved into the National Park we found only a tendency of increasing speeds with increasing distance travelled along the fence. The GLMM on Gamma-distributed response data and logarithmic link function with distances travelled along fence (95%) as predictor, species as well as crossing direction as interaction terms and individual ID as a random factor was superior to both, the full model which additionally contained the hole ID as categorical predictor (delta AIC > 2) and the null model containing only the random factor (ANOVA: Chi-sqaured: 105 on 2 degrees of freedom, p < 0.001). The estimate for springbok moving out of Etosha National Park was 0.64 ± 0.08 SD (t-value: 7.8 , p < 0.001) and for eland 0.64 ± 0.08 SD (t-value: 8.4, p < 0.001). Marginal Pseudo-R² calculated with trigamma method was 0.18 (r.squaredGLMM-function from MuMIn-package, conditional: 0.22).

Results for analysis of average speeds after fence crossing related to the distance travelled along the fence. Average speeds (single observations as colored points, color indicates individual) after a fence interaction (averahe of all speeds 45 min after the fence interaction) on y-axis. Total distance travelled along the fence-line on x-axis. Data is limited to average distances travelled along fence lines (Upper boundary of 100% CI for means, calculated species wise by bootstrapping with 5000 runs). Prediction of GLMM indicated by black line with 95% CI shown by grey shaded area.

Figure 11: Results for analysis of average speeds after fence crossing related to the distance travelled along the fence. Average speeds (single observations as colored points, color indicates individual) after a fence interaction (averahe of all speeds 45 min after the fence interaction) on y-axis. Total distance travelled along the fence-line on x-axis. Data is limited to average distances travelled along fence lines (Upper boundary of 100% CI for means, calculated species wise by bootstrapping with 5000 runs). Prediction of GLMM indicated by black line with 95% CI shown by grey shaded area.

Discussion

The identified fence interactions of the three African antelopes show that animal-fence interactions are commonplace in the study area and affect the ungulates locomotion directly and, in parts, tremendously. Our dataset, consisting out of 2.1 million GPS positions from 25 individuals and representing movements over a period of two and a half years, enabled us to record more than 2,500 km of movement tracks along fence lines. More than the half of this distance was travelled by the nomadic eland (seven individuals), while partial migratory springbok (ten individuals) travelled a third and sedentary kudu (8 individuals) a fifth of this distance along fences. With records of animal-fence interaction on 74 % of all of our tracking days, which is found in other studies as well (Laguna et al. (2022)), we see these to be the rule rather than the exception. Further, animals spend substantial amounts of time at fence-lines and frequently increased speeds if they encountered the veterinary cordon fence (VCF), which had breaches mainly caused by elephants. Such findings demonstrate that fences, despite their semi-permeability, can directly affect the locomotion of ungulates and may ultimately cause energy loss in wildlife. Although our results are limited in their spatial extent and the spectrum of ungulate species, the amount of observations and the level of detail in combination with findings of other studies allow to derive insights for future fence-related research and future fence management planning.

Fence interaction typing

The majority of fence interactions from the rather migratory species springbok and eland was quick, which aligns well with other studies analysing GPS-tracking based animal-fence interactions (Xu et al. (2021); Laguna et al. (2022)). However, this was not the case for sedentary kudu, which dominantly spend longer times at fences either during stay or trace interactions. Further, based on our high-frequency GPS-record schedule (5 - 15 minutes) we identified relatively more trace events than studies using hourly localizations, displaying the limitations and possibilities of movement track analysis in relation to the temporal resolution of the data (Nathan et al. (2022)). Since e.g. trace cross interactions - regardless of the species - lasted 22 minutes, 95% CI [20, 24] on average, they would have been classified differently or not at all, if the temporal resolution of the tracking data would have been coarser than 15 minutes. Notably, stay cross fence interactions, which we classified as such if they consisted out of at least one sequence (minimum of 15 minutes) of low speeds (below 0.05 m/s), lasted 159 minutes, 95% CI [143, 176] on average. Within such events the animals stood on average 68 minutes, 95% CI [57, 79] at fences. Consequently, identified trace interactions from studies using hourly localization may rather display actual events of staying at fences than purely the tracing of fence lines. Already this example demonstrates that a systematic review of animal-fence interaction typing with regard to the temporal resolution of the tracking data and a consistent and appropriate nomenclature is needed to further improve comparability of studies on animal-fence interactions.

Many studies show that linear, man-made landscape features, such as roads and fences, affect the movements of terrestrial animals (see McInturff et al. 2020 and cited literature therein) and negatively influence space use (e.g. Robb et al. 2022). Clearly, individual movement types and the density and permeability of such features within the used space determine the severity of such effects thereby. This general landscape configuration - movement type relation is well reflected in our counts of animal-fence interaction on the different fence types. While most of the migratory springbok and all of the nomadic eland frequently crossed the permeable VCF, only a minority of sedentary kudu did so. Home ranges of these kudu individuals contained the VCF and its permeability enabled the home range formation around it, leading to frequent interactions. We found comparably high numbers of interactions of many kudu with the non-permeable electrified fence (EF). Those individuals’ home ranges were bordered by the fence and home range formation was likely limited by it. In contrast, springbok and eland interacted rather seldom with the EF. We see two main aspects to cause this finding. First, the general movement behaviour of springbok and eland likely enabled them to discard areas beyond the EF from their home ranges and rather select for areas around and beyond the permeable VCF. Indeed, e.g., Robb et al. (2022) show that non-permeable features are strongly avoided by migratory pronghorn (Antilocapra americana). Whereas kudus’ range residency and movement - feeding habits (De Garine-Wichatitsky et al. (2004)) likely urged the individuals to move to the EF regularly. Second, vegetation beyond the EF is rather dominated by woody species and therewith more attractive for browsing kudu, while vegetation beyond the VCF is more open and grass dominated and therewith more attractive for mixed-feeding (and in green season grazing) springbok and eland. This might urged kudu individuals whose home range was close to the EF more towards it.

Overall our procedure of animal-fence interaction typing, which is comparable to those of Xu et al. (2021) and Laguna et al. (2022), showed that already in terms of counts differences between animals of diverse movement type exist. Further it indicates that permeability itself and its consistency is crucial in terms of the types animals interact with fences (see also e.g., Dupuis-Desormeaux et al. 2018; Segar and Keane 2020).

Fence interaction timing

We found clear seasonal trends in numbers of fence interaction per 60-day periods. Peaks occurred during times of seasonal shifts, which can be attributed to changes in the importance of resources in combination with the spatial arrangement of the fences. During colder dry season with limited availability of forage the access to water is crucial to minimize the consequences of adaptations to dehydration such as increased digestion periods and reduced metabolic rates ((Cain2006?)). During warmer green season and during the lactation period the access to productive foraging grounds is essential. On the one hand, the private reserve south of the VCF maintains artificial water points at a high spatial density. On the other hand a rainfall and soil moisture gradient stretches from south-west to north-east in the study region and the VCF runs across that gradient from west to east. So that during the transition from dry to green season (late dry season - early green season), when first flushes of woody plants form a viable and reliable source of nutrition (Archibald and Scholes 2007; Ibrahim et al. 2021) but individuals still drank at water points, they often commuted between vital forage patches north of the VCF and the water points south of it. Overall this caused the observed peaks in visited above ground woody biomass and in animal-fence interactions. Further studies on animal-fence interactions find seasonal differences in the occurrence of such (e.g., Wilkinson et al. 2021; Xu et al. 2021) and demonstrate the importance of the spatial arrangement of fences in relation to movement strategies (Jones et al. 2019). With that we see an urgent need for animal movement modelling studies, which evaluate best practices for fence arrangements, e.g., along important environmental gradients, as to minimize possible negative effects of fences on terrestrial animals. Moreover, these findings demonstrate that the severity of direct effects of fences on animal movement vary across time and possible measures to mitigate these must take such seasonal differences into consideration.

Daytimes of animal-fence interactions clearly depend on species activity patterns. For instance, at Lake Nakuru National Park, Kenya, Wilkinson et al. (2021) found that ungulates, which exhibit main activity times during the day [e.g.; N. Owen-Smith and Goodall (2014)], were more likely to cross the fence during the day (defined as fixed times). Also we found the majority of fence interactions to occur 70 - 160 minutes after sunrise and could add species-wise peaks in relation to sunrise to this knowledge. Comparing the occurrence of fence interactions with times of directed movements (relative turning angles below 90 degrees, speeds larger than 1.8 m/s) in relation to daytimes, we found that they coincide. This indicates that fences impede ungulates when they move towards a certain location and that major time windows of directed movements exist (e.g., N. Owen-Smith, Fryxell, and Merrill 2010). These are important aspects to be considered, when fencing is intended to be more “wildlife-friendly” Segar and Keane (2020) and fence passages might be implemented strategically (e.g., Dupuis-Desormeaux et al. 2018). For example, optimal inter-passage distance could be derived from movement data as the distance a species could easily walk during its major times of directed movements. With major times of springboks’ directed movements ranging from one to seven hours after sunrise and average speeds being 0.177 m/s, 95% CI [.175, 0.180] during late dry season, this would result in a may optimal inter-passage distance of 3.8 km for our situation. Clearly, further studies would be needed to evaluate a standardized procedure for this and guide future fence and fence-gap planning.

Fence interaction relating to movement type

We did not find an indication for a correlation between individual space-use (area) and the intensity of animal-fence interaction (measured as distance or time along fences)

Fence effects on locomotion

The movement tracks of three large African antelope species highlight the importance of the VCFs’ permeability to avoid direct energy loss at the fence and to access sufficient forage under temporally and spatially erratic plant production. Based on our high-frequency movement data of 24 GPS-collared individuals with a total 1.7 million GPS localizations, we identified thousands of fence interactions over a period of 27 months. On 68 % of all of our tracking days we recorded interactions of the tracked individuals with the VCF. The majority of these interactions (88 %) was observed for migratory springbok and eland. Both species often travelled long distances along the VCF (on average springbok: 28.9 ± 11.2 (SD) km / year, eland: 64.5 ± 52.2 (SD) km / year).

We classified the fence interactions into different interaction types similar to the approach of Xu et al. (2021) and were able to identify an additional interaction type stay, due to the high temporal resolution of the GPS localizations (5-15 min). For each interaction, the assigned energetic signatures showed that depending on the type of interaction, antelopes either spend less, the same amount, or in many cases more energy when they interact with a fence than immediately before or after encountering the fence. Interestingly, when we compared fence interactions of the same type but with differing crossing success, we found an increased energy expenditure when crossing was successful. Reviewing the 21 studies identified to consider fence related energy expenditure in McInturff et al. (2020), we did not find a single one to include direct measurements of ODBA at fences. Rather, the studies examined behavior through observations and focussed on the effectiveness of fencing. In this sense our findings fundamentally contribute to the emerging field of fence ecology by providing first insights to the energetic impacts of fences for African ungulates.

The individuals which successfully crossed the VCF were capable of following patchy distributed feeding-resources over a large spatial scale (average area [95 % utilisation distribution] used at ENP or beyond the VCF: springbok: 246 ± 493 (SD) km², eland: 193 ± 217 (SD) km²) and benefited from a clear gain in forage quality and quantity. We found temporal peaks in fence crossing shortly before the onset of the growing season and, connected to that, a significant increase in crossing probability shortly before the onset of the first seasonal rain. The ungulates visited greener patches (11.9 % greener on average) and also expanded the duration of their experienced green season as they reached some patches with early occurring first flushes and others with delayed dry out. This is an immanent behaviour of non-sedentary ungulates (e.g., Esmaeili et al. 2021) but demonstrating this in relation to a VCF and for native antelopes of southern Africa has to our knowledge never been done before and provides fundamental knowledge for strategic management planning.

Seasonal dynamics of wildlife-fence interactions

Our findings support evidence that seasonal patterns of plant-phenology trigger large scale movement of non-sedentary ungulates globally (e.g., Norman Owen-Smith et al. (2020) (Africa), Peters et al. (2019) (Europe), Aikens et al. (2020) (North America), Schroeder et al. (2014) (South America), Nandintsetseg et al. (2019) (Eastern Asia)). Our simple model to predict crossing events in relation to time towards the seasonal turning point appears plausible. However, the fact that crossings peaked weeks before we measured the first significant rainfall was unexpected. We see a combination of mechanisms driving this finding.

In contrast to temperate climates where greening occurs along temperature gradients in a wave-like pattern (Aikens et al. 2020), the greening in most semi-arid landscapes strongly depends on effective rainfall (Archibald and Scholes 2007; Ibrahim et al. 2021). In combination with an uneven spatial distribution of such rainfall events, this creates an erratic mosaic of greening patches along ephemeral rivers, in depressions and at the rainfall locations themselves at different times. While herbaceous plants respond directly to rainwater in the top soil layer, woody plants can store water and many trees and shrubs flower and foliate weeks before the rain season starts (Archibald and Scholes 2007; Ibrahim et al. 2021). These first flushes are of good nutritional value (Makhado, Potgieter, and Luus-Powell 2016; Marius et al. 2021) and are eaten by many ungulates, irrespective of the feeding type (Skinner and Chimimba 2005). Consequently, ungulates will track patches of sufficient flushed biomass in semi-arid landscapes. Notably, they must be capable of reaching such greened places rapidly, since growth periods of arid-adapted vegetation are often short (Noy-Meir 1973). On the one hand, large depressions such as the Makgadikgadi pans in Botswana or the Etosha pan in Namibia form reliable grazing sources as water from vast and wetter areas drains into them, which causes periodical ungulate migrations like the Makgadikgadi zebra migration (Bartlam-Brooks, Bonyongo, and Harris 2011) or the historic West - East Etosha migration of oryx and elephant (Tinley 1971). On the other hand, patches, which green-up due to local rainfall events, drive the movements of ungulates at a smaller scale. For instance, all of the tracked springbok groups moved northwards across the VCF to a drainage depression immediately after the first rainfalls in 2019 when a severe drought ended, but did not move there the following two green seasons. In this sense, small scale rainfall events might have triggered some antelope movements across the fence before we measured the first rainfall of the respective year.

The detected peak in fence crossings towards the end of the dry season is likely to be caused by local green-ups across the fence where the ungulates moved to. Simultaneously, as natural surface water was not available and the overall green biomass was still low, the animals regularly moved back to the artificial water holes within the private reserve and, thereby, crossed the fence very often. Interestingly, once the rain seasons, started some individuals went long distances and times outside the private reserve (up to 80 km and more than 6 months away), others did shorter excursions outside and yet others did not leave the private reserve at all. When vegetation greenness dropped again, the tagged individuals of all ungulate species returned to the private reserve and visited the artificial waterholes on almost a daily basis. These dynamics show that fence interactions occur all year long, but depending on movement types seasonal peaks appear. At least for the VCF, where gaps were created by elephants, we found a drop in fence encounters during the green season caused by the ability of many individuals to occupy sufficient feeding areas regardless of the fence side.

Our results confirm recent findings of Wilkinson et al. (2021), who found ungulates to mostly cross fences during the day but did not find the seasons to influence this behavior. Focusing on crossing per se, we found clear seasonal dynamics in interactions with the VCF, demonstrating that only during peak dry seasons the ungulates’ movements were less affected by this fence, while an intact VCF would have blocked them during the rest of the year. Conclusively, a permanent non-permeable fence would have tremendous fitness consequences for antelopes, especially when the green season is about to start.

Direct effects of wildlife-fence interactions on energy expenditure

We found major forms of animal-fence interactions to be either quick, trace-like or including a stay event. This contrasts with the findings of Xu et al. (2021) who found Odocoileus hemionus (mule deer) and Antilocapra americana (pronghorn) to dominantly interact quickly with fences. These differences likely result from the lower temporal GPS resolution they used (hourly vs. 5 - 15 minutes in our study). The high temporal resolution of our localization data enabled us to identify hundreds of events during which the antelopes paused at fences for considerable time. This finding adds to the types of animal-fence interactions identified by Xu et al. (2021) (based on movement data) and is in line with the frequently observed “vigilance” behavior of Wilkinson et al. (2021) at Lake Nakuru National Park (Kenya) (based on camera trap data). Of our 24 tracked individuals, 18 crossed the VCF more than 1400 times in total within the two years of observation. We recorded crossings for all three species, with eland and springbok regularly trespassing the VCF. Permeable fences are often crossed by ungulates and other wildlife (e.g., Kesch, Bauer, and Loveridge 2014; Dupuis-Desormeaux et al. 2018; Jones et al. 2019; Seigle‐Ferrand et al. 2021), although their movement patterns are affected (e.g., McKillop and Sibly 1988; Harrington and Conover 2006; Sawyer et al. 2013; Laskin et al. 2020; Wilkinson et al. 2021). Permeable fences may not act as a general barrier but as an obstacle which leads to altered behavior and, consequently, energy expenditure. The latter we explicitly analyzed through measured changes in ODBA of the animal and found direct effects of the fences on the antelopes’ motion patterns.

We identified plenty of events with eland or kudu spending dozens of minutes to a few hours at the same location while encountering a fence (stay non-cross). Comparably low ODBA proposes stationary halts or vigilance behavior. This may indicate confusion of the antelopes when a fence blocked their intended movement, followed by a new movement decision-making process. Such fence interactions lasted on average more than 2 hours and are frequently observed in ungulates (Wilkinson et al. 2021; Park et al. 2021). Although not directly spending more energy, ungulates likely lose a lot of time during such events. For instance, we found an eland to spend a total of 4 days out of the 442 observation days during such stay non-cross fence interactions (32 single events).

Most fence interactions resulted in increased ODBA (most frequent consequence of interaction in eland, second most in springbok and kudu). This highlights the severe direct impact fences can have on energy expenditure of wildlife. In cases of crossing at the VCF (quick cross of eland), the high ODBA values likely result from the necessity to jump over half broken fence sections, entanglement in wires or similar behavior as observed for many ungulate species globally (e.g., Harrington and Conover 2006; Laskin et al. 2020; Wilkinson et al. 2021). In cases of non-cross (quick non-cross of springbok, trace non-cross of springbok and kudu), the increased activity at the fences is likely caused by attempts to cross or urgent behavior during the search for a gap. Ungulates will cross (semi-) permeable fences and will frequently try to cross impermeable fences which traverse their habitats. For instance, Xu et al. (2021) estimate 250 fence encounters per year for a pronghorn migrating through the rangelands of western Wyoming, USA; we found 44 encounters for a springbok only at the VCF (maximum: springbok 6770 with 134 encounters in 2020). Bearing this in mind, single short-term events of increased energy expenditure will sum up over time and, thus, negatively influence individual long-term energy balances.

Further, we identified many situations without any significant change of ODBA at the fence. Despite non-cross interactions of trace - like walks along the fence line, which might represent only slight changes in movement direction due to the barrier or periods of foraging events, we found specific types of cross events in all species to not clearly affect ODBA. Likely these events occurred on locations with more or less clear, easily identifiable gaps and persistent gaps allowing successful crossing at known locations and thus becoming a habituated behavior, which has been documented earlier (e.g., Dupuis-Desormeaux et al. 2018; Wilkinson et al. 2021).

These first results on energetic consequences of fences for ungulates demonstrate that, first, impermeable fence lines will lead to an increase in motion and thus in energy expenditure. Second, they show that inconsistent and vague permeable fence lines will lead to high energy expenditure during the crossing. Third, persistent and obvious gaps will not directly affect energy expenditure during crossing. Therefore, our findings form a solid base for further investigations on such short-term effects, e.g., the temporal dynamics of energy expenditure along each interaction event or the long-term dynamics at specific locations as to study learning effects.

Indirect effects of fence gaps on energetics

Individuals of all three antelope species which travelled beyond the VCF fed on up to 40 % greener vegetation as compared to the average found in their home range. They were also able to move to areas with fresher biomass and, thus, optimized energy intake. This emphasizes the importance of permeable fences for ungulates that enable them to track a sufficient amount of resources.

Our results align well with recent findings on the forage maturation hypothesis (Esmaeili et al. 2021) and demonstrate the dynamics in animal habitat formations (Aikens et al. 2020). Esmaeili et al. (2021) extended Fryxell (1991) and others’ work where highest forage quality occurs at intermediate maturation stages of plants. They adjusted the maturation stage - energy uptake optima according to body sizes, digestive system and water dependency of larger herbivores and tested for this using a global set of animal movement and plant phenology data. Although they generally show that the smaller the herbivores are the more they select for high quality forage, they particularly did not find springbok to select for high quality forage and slightly avoiding higher quantities. Already in the postulation of the forage maturation hypothesis, Fryxell (1991) points out that forage maturation effects should be limited under semi-arid conditions, with short and unpredictable rainfall, and in smaller, selective ruminants. However, Esmaeili et al. (2021) worked on relatively coarse temporal and spatial scales which might have concealed species-specific, dynamic selectivity for forage quantity and quality. As claimed by Nghiyalwa et al. (2021), remote sensing based analysis of phenology dynamics in savannas requires high resolution satellite data. By using such and focussing on vegetation types our results show dynamic patterns of forage ground selection. For instance, we found springbok during the beginning of the green season to be on patches of higher greenness (as compared to the average of all patches of a vegetation type within each individuals home range) but mostly not of higher greening, while selected patches during peak and end of the green season showed rather higher greening rates than greenness, pointing at the selectivity of these rather small ungulates to high-quality forage. We found the large eland to constantly be on patches of above-average greenness with fluctuating greening, demonstrating their selectivity for patches of high quantities of forage.

Aikens et al. (2020) demonstrate that habitat formation in ungulates is a dynamic process following the fluxes of resource availability. They focus on temperate landscapes with wave-like green-up patterns. However, the principle that ungulates move to areas where at a specific point in time high-quality forage is available must hold true for other climates too. Indeed, it has been known for decades that many southern African ungulates switch food sources during the course of the year (Skinner and Chimimba 2005). Our analysis shows that the tracked species move to specific vegetation patches of better phenological quality than others within their home range. With this, we provide a basis for further studies in order to expand the green-wave surfing concept to dynamic habitat formation (Aikens et al. 2020) and show in detail how remarkably these animals track the scattered resources of such semi-arid landscapes, a fact that has been known to indigenous peoples since antiquity (Roche 2005).

Management implications and future perspective

We see gaps in the VCF, which were mostly created by elephants, to be of significant benefit for our study antelopes. Especially the mixed-feeding springbok and eland were able to track scattered, high-quality resources on a large scale and thus likely improved individual health. Furthermore, these gaps mediated the grazing pressure between the pastures on both sides of the fence and, thereby, likely reduced the risk of ecosystem degradation. Also, as ungulates were able to reach resources and mates on both fence sides, these gaps likely play a key role in improving genetic diversity and reproductive success. In the future, the larger herbivores will have an exacerbated need to reach resource patches across vast areas since the degree of aridity in south west Africa will increase as a result of climate change (Iturbide et al. 2020) and suitable habitats are likely to move beyond the boundaries of current protected areas, increasing the need for larger unfragmented areas (Turpie et al. 2010).

Although this study focuses on interactions between wild herbivores and the VCF, we recognize that the VCF plays a major role in Namibia’s beef economy due to international requirements for certified disease-free meat exports, which include the presence of an effective fence. However, our findings could add to the ongoing debate regarding social, economic and ecological benefits and impacts of this fence (e.g., Sutmoller 2002; Scoones et al. 2010; Weaver et al. 2013; Mogotsi, Kgosikoma, and Lubinda 2016). Within this debate, the benefits and the long-term maintenance of the VCF are increasingly being questioned. While negative effects on wildlife ecology are observed (Martin 2005), the trade regulations set by the World Organization for Animal Health (OIE) induce the government to maintain the fence, which is frequently damaged by elephants and other wildlife.

A key problem with gaps created by elephants is that animals which use these gaps, such as our study ungulates and other species, are exposed to a severe risk of entanglement. Temporary repairs may only increase this risk since animals are habituated to use such sections and will likely intensively try to cross at these locations. Consequently, effective measures are needed to eliminate such injury risks in the future.

A first step could be the implementation of managed gaps in the VCF on locations where complimentary land-uses on both sides of the fence minimize the risk of wildlife-livestock disease transfer. Many studies have shown that there are a variety of effective options for such. Gaps can be installed as to be selective for certain species (e.g., exclusion of rhinos by rock walls: Dupuis-Désormeaux et al. (2016); exclusion of livestock by tubular metal grills: Gates et al. (2012); selective fence designs: Laskin et al. (2020); Segar and Keane (2020)) or simply open for all species. Animals will incorporate them into their movement corridors over time (Dupuis-Desormeaux et al. 2018). Since our study recognized clear temporal peaks in cross-fence movements, in the event of permanent corridors not being an option, a gated system at pressure points in the VCF could be opened at peak movement periods.

Important opportunities would arise from the implementation of gaps as a management tool. Since technical solutions for an automated surveillance of wildlife exist (e.g., semi-automated species identification of camera trap data: Janzen et al. (2017)), and even individuals of certain species can be identified (e.g., zebras: Gosling L. M., Muntifering J., Kolberg H., Uiseb K. (2019)), populations could be monitored at the gaps in fences. Simultaneously, the health status of individuals could be checked and, in times of disease outbreaks, gaps could be closed to restore an impermeable barrier.

A major challenge thereby is the placement of gaps. Clearly, the better they are located on natural movement corridors, the less animals will be restricted. Our data shows that gaps created by elephants were intensively used, especially on either ephemeral rivers or where the fence cuts off larger patches of plant communities. This implies that elephants break fences on important locations naturally; we see these gaps as ‘animal-informed movement corridors’ which likely form a solid baseline for placement decisions.

Conclusion

High resolution tracking, accelerometer and satellite imagery data enabled us to uncover new insights into ungulate-fence interactions and connected consequences. Along the 70 km section of the VCF between ENP and EH, we documented 1471 crossings of springbok, kudu and eland (17 individuals in total) during the 27 month period. These were of a total of 2905 fence interactions (at the VCF and other fences) mostly resulting in augmented behavior by the study animals, which was often related to an increase in energy expenditure when encountering a fence. Particularly, non-sedentary eland and springbok were mainly affected by the VCF, while sedentary kudu were rarely affected. Fence interactions showed seasonal dynamics, peaking right before first seasonal rains. Gaps within the VCF allowed individuals to cross in order to track scattered resources on vast areas, leading to foraging benefits. This would have been impeded if the fence was an impermeable barrier.

Antelopes inhabiting semi-arid environments face a patchy distribution of resources in both time and space. Depending on foraging and movement type, they conquer this sparsity by travelling through vast areas in an adaptive, resource-tracking manner. Wildlife-proof fences stop such movements and, thereby, tremendously interfere with connected ecological processes. The latter can only be restored or sustained if fences are either permeable, e.g., by containing gaps, or are removed completely.

Our findings provide a solid base for management decisions and future research. We emphasize that research focussing on resource-tracking of ungulates in semi-arid ecosystems must use high spatial resolution satellite data of plant phenology, ideally vegetation-type specific, to account for the patchy distribution of resources and to identify small-scale reactions. Regarding the VCF’s future, we see managed gaps to be a good first step towards a sustainable development of these unique, native ecosystems.

Acknowledgments

We thank Andre Nel and the team of Etosha Heights Private Reserve for full support during field research. We acknowledge the support of the Ministry of Environment, Tourism and Forestry, Namibia and the Namibian National Commission On Research Science & Technology, who permitted this research (certifcate number RCIV00032018 with authorisation numbers: 20190602, 20190808). Funding: This work was part of the ORYCS project within the SPACES II program, supported by the German Federal Ministry of Education and Research [grant number FKZ 01LL1804A].

References

Aikens, Ellen O., Atle Mysterud, Jerod A. Merkle, Francesca Cagnacci, Inger Maren Rivrud, Mark Hebblewhite, Mark A. Hurley, et al. 2020. Wave-like Patterns of Plant Phenology Determine Ungulate Movement Tactics.” Current Biology 30 (17): 3444–3449.e4. https://doi.org/10.1016/j.cub.2020.06.032.
Archibald, S., and R. J. Scholes. 2007. Leaf green-up in a semi-arid African savanna -separating tree and grass responses to environmental cues.” Journal of Vegetation Science 18 (4): 583–94. https://doi.org/10.1111/J.1654-1103.2007.TB02572.X.
Bartlam-Brooks, H. L. A., M. C. Bonyongo, and Stephen Harris. 2011. Will reconnecting ecosystems allow long-distance mammal migrations to resume? A case study of a zebra Equus burchelli migration in Botswana.” ORYX 45 (2): 210–16. https://doi.org/10.1017/S0030605310000414.
Bigalke, R. C. 1972. Observations on the Behaviour and Feeding Habits of the Springbok, Antidorcas Marsupialis.” Zoologica Africana 7 (1): 333–59. https://doi.org/10.1080/00445096.1972.11447448.
Bunnefeld, Nils, Luca Börger, Bram Van Moorter, Christer M. Rolandsen, Holger Dettki, Erling Johan Solberg, and Göran Ericsson. 2011. A model-driven approach to quantify migration patterns: Individual, regional and yearly differences.” Journal of Animal Ecology 80 (2): 466–76. https://doi.org/10.1111/j.1365-2656.2010.01776.x.
Burkholder, Emily N, Andrew F Jakes, Paul F Jones, Mark Hebblewhite, and Chad J Bishop. 2018. To jump or not to jump: Mule deer and white-tailed deer fence crossing decisions.” Wildlife Society Bulletin 42 (3): 420–29. https://doi.org/10.1002/wsb.898.
Caron, A., E. Miguel, C. Gomo, P. Makaya, D. M. Pfukenyi, C. Foggin, T. Hove, and M. De Garine-Wichatitsky. 2013. Relationship between burden of infection in ungulate populations and wildlife/livestock interfaces.” Epidemiology and Infection 141 (7): 1522–35. https://doi.org/10.1017/S0950268813000204.
Cumming, D. H. M., S. A. Osofsky, S. J. Atkinson, and M. W. Atkinson. 2015. Beyond fences: wildlife, livestock and land use in Southern Africa. In One Health: The Theory and Practice of Integrated Health Approaches, 243–57. CABI. https://doi.org/10.1079/9781780643410.0243.
De Garine-Wichatitsky, Michel, Hervé Fritz, Iain J. Gordon, and Andrew W. Illius. 2004. Bush selection along foraging pathways by sympatric impala and greater kudu.” Oecologia 141 (1): 66–75. https://doi.org/10.1007/s00442-004-1630-3.
Dupuis-Desormeaux, Marc, Zeke Davidson, Laura Pratt, Mary Mwololo, and Suzanne E. MacDonald. 2016. Testing the effects of perimeter fencing and elephant exclosures on lion predation patterns in a Kenyan wildlife conservancy.” PeerJ 2016 (2): e1681. https://doi.org/10.7717/peerj.1681.
Dupuis-Desormeaux, Marc, Timothy N. Kaaria, Mary Mwololo, Zeke Davidson, and Suzanne E. MacDonald. 2018. A ghost fence-gap: Surprising wildlife usage of an obsolete fence crossing.” PeerJ 2018 (11): e5950. https://doi.org/10.7717/peerj.5950.
Dupuis-Désormeaux, Marc, Zeke Davidson, Mary Mwololo, Edwin Kisio, and Suzanne E. MacDonald. 2016. Usage of specialized fence-gaps in a black rhinoceros conservancy in Kenya.” African Journal of Wildlife Research 46 (1): 22–32. https://doi.org/10.3957/056.046.0022.
Esmaeili, Saeideh, Brett R. Jesmer, Shannon E. Albeke, Ellen O. Aikens, Kathryn A. Schoenecker, Sarah R. B. King, Briana Abrahms, et al. 2021. Body size and digestive system shape resource selection by ungulates: A cross‐taxa test of the forage maturation hypothesis.” Edited by Jean‐Michel Gaillard. Ecology Letters 00 (July): ele.13848. https://doi.org/10.1111/ele.13848.
Fryxell, J. M. 1991. Forage quality and aggregation by large herbivores.” American Naturalist 138 (2): 478–98. https://doi.org/10.1086/285227.
Funk, Chris, Pete Peterson, Martin Landsfeld, Diego Pedreros, James Verdin, Shraddhanand Shukla, Gregory Husak, et al. 2015. The climate hazards infrared precipitation with stations - A new environmental record for monitoring extremes.” Scientific Data 2 (1): 1–21. https://doi.org/10.1038/sdata.2015.66.
Gates, C. Cormack, Paul Jones, Michael Suitor, Andrew Jakes, Mark S. Boyce, Kyran Kunkel, and Kevin Wilson. 2012. The influence of land use and fences on habitat effectiveness, movements and distribution of pronghorn in the grasslands of North America.” In Fencing for Conservation: Restriction of Evolutionary Potential or a Riposte to Threatening Processes?, 277–94. Springer New York. https://doi.org/10.1007/978-1-4614-0902-1_15.
Goddard, P. J., R. W. Summers, A. J. Macdonald, C. Murray, and A. R. Fawcett. 2001. Behavioural responses of red deer to fences of five different designs.” Applied Animal Behaviour Science 73 (4): 289–98. https://doi.org/10.1016/S0168-1591(01)00150-2.
Gosling L. M., Muntifering J., Kolberg H., Uiseb K., King S. R. B. 2019. Equus zebra ssp. hartmannae.” IUCN Red List of Threatened Species. 8235. https://doi.org/https://dx.doi.org/10.2305/IUCN.UK.2019-1.RLTS.T7958A45171819.en.
Gulsby, William D., Daniel W. Stull, George R. Gallagher, David A. Osborn, Robert J. Warren, Karl V. Miller, and Lawrence V. Tannenbaum. 2011. Movements and home ranges of white-tailed deer in response to roadside fences.” Wildlife Society Bulletin 35 (3): 282–90. https://doi.org/10.1002/WSB.38.
Harrington, Justin L., and Michael R. Conover. 2006. Characteristics of Ungulate Behavior and Mortality Associated with Wire Fences.” Wildlife Society Bulletin 34 (5): 1295–305. https://doi.org/10.2193/0091-7648(2006)34[1295:coubam]2.0.co;2.
Hayward, Matthew W. 2012. Perspectives on fencing for conservation based on four case studies: Marsupial conservation in Australian forests; bushmeat hunting in South Africa; large predator reintroduction in South Africa; and large mammal conservation in Poland.” In Fencing for Conservation: Restriction of Evolutionary Potential or a Riposte to Threatening Processes?, 7–20. Springer New York. https://doi.org/10.1007/978-1-4614-0902-1_2.
Hempson, Gareth P., Sally Archibald, and William J. Bond. 2015. A continent-wide assessment of the form and intensity of large mammal herbivory in Africa.” Science 350 (6264): 1056–61. https://doi.org/10.1126/science.aac7978.
Hillman, Jessec. 1988. Home range and movement of the common eland (Taurotragus oryx Pallas 1766) in Kenya.” African Journal of Ecology 26 (2): 135–48. https://doi.org/10.1111/j.1365-2028.1988.tb00964.x.
Ibrahim, Sa’Ad, Jörg Kaduk, Kevin Tansey, Heiko Balzter, and Umar Mohammed Lawal. 2021. Detecting phenological changes in plant functional types over West African savannah dominated landscape.” International Journal of Remote Sensing 42 (2): 567–94. https://doi.org/10.1080/01431161.2020.1811914.
Iturbide, Maialen, José M. Gutiérrez, Lincoln M. Alves, Joaquín Bedia, Ruth Cerezo-Mota, Ezequiel Cimadevilla, Antonio S. Cofiño, et al. 2020. An update of IPCC climate reference regions for subcontinental analysis of climate model data: definition and aggregated datasets.” Earth System Science Data 12 (4): 2959–70. https://doi.org/10.5194/ESSD-12-2959-2020.
Jakes, Andrew F., Paul F. Jones, L. Christine Paige, Renee G. Seidler, and Marcel P. Huijser. 2018. A fence runs through it: A call for greater attention to the influence of fences on wildlife and ecosystems.” Elsevier Ltd. https://doi.org/10.1016/j.biocon.2018.09.026.
Janzen, Michael, Kaitlyn Visser, Darcy Visscher, Ian MacLeod, Dragomir Vujnovic, and Ksenija Vujnovic. 2017. Semi-automated camera trap image processing for the detection of ungulate fence crossing events.” Environmental Monitoring and Assessment 189 (10): 1–13. https://doi.org/10.1007/s10661-017-6206-x.
Jones, Paul F., Andrew F. Jakes, Andrew C. Telander, Hall Sawyer, Brian H. Martin, and Mark Hebblewhite. 2019. Fences reduce habitat for a partially migratory ungulate in the northern sagebrush steppe.” Ecosphere 10 (7). https://doi.org/10.1002/ecs2.2791.
Kauffman, Matthew J., Francesca Cagnacci, Simon Chamaillé-Jammes, Mark Hebblewhite, J. Grant C. Hopcraft, Jerod A. Merkle, Thomas Mueller, et al. 2021b. Mapping out a future for ungulate migrations.” Science 372 (6542): 566–69. https://doi.org/10.1126/science.abf0998.
———, et al. 2021a. Mapping out a future for ungulate migrations.” Science 372 (6542): 566–69. https://doi.org/10.1126/science.abf0998.
Kesch, Kristina M., Dominik T. Bauer, and Andrew J. Loveridge. 2014. Undermining game fences: Who is digging holes in Kalahari sands? African Journal of Ecology 52 (2): 144–50. https://doi.org/10.1111/aje.12096.
Laguna, Eduardo, José A. Barasona, Antonio J. Carpio, Joaquín Vicente, and Pelayo Acevedo. 2022. Permeability of artificial barriers (fences) for wild boar (Sus scrofa) in Mediterranean mixed landscapes.” Pest Management Science. https://doi.org/10.1002/ps.6853.
Laskin, David N., Dillon Watt, Jesse Whittington, and Karsten Heuer. 2020. Designing a fence that enables free passage of wildlife while containing reintroduced bison: A multispecies evaluation.” Wildlife Biology 2020 (4). https://doi.org/10.2981/wlb.00751.
Lindsey, P A, S. S. Romañach, and H T Davies-Mostert. 2009. The importance of conservancies for enhancing the value of game ranch land for large mammal conservation in southern Africa.” John Wiley & Sons, Ltd. https://doi.org/10.1111/j.1469-7998.2008.00529.x.
Lindsey, Peter A. 2011. An Analysis of Game Meat Production in Namibia: Linkages with Food Security.
Makhado, R. A., M. J. Potgieter, and W. J. Luus-Powell. 2016. Nutritional value of colophospermum mopane as source of browse and its chemical defences against browsers: A review.” Journal of Animal and Plant Sciences 26 (3): 569–76.
Marius, Lucia N., Maria N. T. Shipandeni, Luis A. Rodríguez-Campos, Emmanuel L. K. Osafo, Irvin D. T. Mpofu, Terry Ansah, Katrina L. Shiningavamwe, Victoria Attoh-Kotoku, and Christopher Antwi. 2021. Seasonal variation in chemical composition and in-vitro gas production of woody plant species of semi-arid condition of Namibia.” Agroforestry Systems 95 (6): 1191–1204. https://doi.org/10.1007/S10457-021-00643-5/TABLES/5.
Martin, Rowan B. 2005. The influence of veterinary control fences on certain wild large mammal species in the Caprivi, Namibia. Occasional Papers of the IUCN Species Survival Commission 30: 27–xxi.
Mbaiwa, B Y Joseph E, and Onaletshepho I Mbaiwa. 2006. The Effects of Veterinary Fences on Wildlife Populations in Okavango Delta, Botswana.” International Journal of Wilderness 12 (3): 17–24.
McInturff, Alex, Wenjing Xu, Christine E. Wilkinson, Nandintsetseg Dejid, and Justin S. Brashares. 2020. Fence Ecology: Frameworks for Understanding the Ecological Effects of Fences.” BioScience 70 (11): 971–85. https://doi.org/10.1093/biosci/biaa103.
McKillop, I. G., and R. M. Sibly. 1988. Animal behaviour at electric fences and the implications for management.” Mammal Review 18 (2): 91–103. https://doi.org/10.1111/j.1365-2907.1988.tb00078.x.
Mogotsi, Kabo, Olaotswe Ernest Kgosikoma, and Keotshepile Fredric Lubinda. 2016. Wildlife-livestock interface, veterinary cordon fence damage, lack of protection zones, livestock theft and owner apathy: Complex socio-ecological dynamics in Foot and Mouth disease control in southern Africa.” Pastoralism 6 (1): 21. https://doi.org/10.1186/s13570-016-0068-7.
Mutinda, Matthew, Geoffrey Chenge, Francis Gakuya, Moses Otiende, Patrick Omondi, Samuel Kasiki, Ramón C. Soriguer, and Samer Alasaad. 2014. Detusking Fence-Breaker Elephants as an Approach in Human-Elephant Conflict Mitigation.” PLOS ONE 9 (3): e91749. https://doi.org/10.1371/JOURNAL.PONE.0091749.
Naidoo, Robin, Piet Beytell, Angela Brennan, Werner Kilian, Graham McCulloch, Amanda Stronza, Russell Taylor, Chief Tsholofelo, and Anna Songhurst. 2022. Challenges to Elephant Connectivity From Border Fences in the World’s Largest Transfrontier Conservation Area.” Frontiers in Conservation Science 3 (February): 14. https://doi.org/10.3389/fcosc.2022.788133.
Nandintsetseg, Dejid, Chloe Bracis, Kirk A Olson, Katrin Böhning-Gaese, Justin M Calabrese, Buyanaa Chimeddorj, | William, et al. 2019. Challenges in the conservation of wide-ranging nomadic species.” J Appl Ecol 56: 1916–26. https://doi.org/10.1111/1365-2664.13380.
Nathan, Ran, Christopher T. Monk, Robert Arlinghaus, Timo Adam, Josep Alós, Michael Assaf, Henrik Baktoft, et al. 2022. Big-data approaches lead to an increased understanding of the ecology of animal movement.” American Association for the Advancement of Science. https://doi.org/10.1126/science.abg1780.
Nghiyalwa, Hilma S., Marcel Urban, Jussi Baade, Izak P. J. Smit, Abel Ramoelo, Buster Mogonong, and Christiane Schmullius. 2021. Spatio-Temporal Mixed Pixel Analysis of Savanna Ecosystems: A Review.” Remote Sensing 2021, Vol. 13, Page 3870 13 (19): 3870. https://doi.org/10.3390/RS13193870.
Noy-Meir, Imanuel. 1973. Desert Ecosystems: Environment and Producers.” Annual Review of Ecology and Systematics 4 (1): 25–51. https://doi.org/10.1146/annurev.es.04.110173.000325.
Owen-Smith, N. 1985. The ecological potential of the kudu for commercial production in savanna regions.” Journal of the Grassland Society of Southern Africa 2 (3): 7–10. https://doi.org/10.1080/02566702.1985.9648005.
Owen-Smith, N., J. M. Fryxell, and E. H. Merrill. 2010. Foraging theory upscaled: The behavioural ecology of herbivore movement.” Royal Society. https://doi.org/10.1098/rstb.2010.0095.
Owen-Smith, N., and V. Goodall. 2014. Coping with savanna seasonality: Comparative daily activity patterns of African ungulates as revealed by GPS telemetry.” Journal of Zoology 293 (3): 181–91. https://doi.org/10.1111/JZO.12132.
Owen-Smith, Norman. 2014. Spatial ecology of large herbivore populations.” Ecography 37 (5): 416–30. https://doi.org/10.1111/j.1600-0587.2013.00613.x.
Owen-Smith, Norman, Grant Hopcraft, Thomas Morrison, Simon Chamaillé-Jammes, Robyn Hetem, Emily Bennitt, and Frank Van Langevelde. 2020. Movement ecology of large herbivores in African savannas: current knowledge and gaps.” Mammal Review, 1–15. https://doi.org/10.1111/mam.12193.
Palmer, William L, Jack M Payne, Robert G Wingard, and John L George. 1985. A Practical Fence to reduce Deer Damage.” Wildlife Society Bulletin 13 (3): 240–45. https://www.jstor.org/stable/3782485?seq=1.
Park, Hee Bok, Donggul Woo, Tae Young Choi, and Sungwon Hong. 2021. Assessment of the Behavioural Response of Korean Water Deer (Hydropotes inermis argyropus) to Different Fence Heights.” Animals 2021, Vol. 11, Page 938 11 (4): 938. https://doi.org/10.3390/ANI11040938.
Peters, Wibke, Mark Hebblewhite, Atle Mysterud, Daniel Eacker, A. J. Mark Hewison, John D. C. Linnell, Stefano Focardi, et al. 2019. Large herbivore migration plasticity along environmental gradients in Europe: life-history traits modulate forage effects.” Oikos 128 (3): 416–29. https://doi.org/10.1111/OIK.05588.
Robb, Benjamin S., Jerod A. Merkle, Hall Sawyer, Jeffrey L. Beck, and Matthew J. Kauffman. 2022. Nowhere to run: semi-permeable barriers affect pronghorn space use.” Journal of Wildlife Management. https://doi.org/10.1002/jwmg.22212.
Roche, Chris. 2005. Springbok ... Drink the Rain’s Blood’: Indigenous Knowledge and Its Use in Environmental History-The Case of the /Xam and an Understanding of Springbok Treks *.” South African Historical Journal 53: 1–22.
Safner, Toni, Ana Gracanin, Ivan Gligora, Boštjan Pokorny, Katarina Flajšman, Marco Apollonio, and Nikica Šprem. 2021. STATE BORDER FENCES AS A THREAT to HABITAT CONNECTIVITY: A CASE STUDY from SOUTH-EASTERN EUROPE.” Sumarski List 145 (5-6): 269–78. https://doi.org/10.31298/SL.145.5-6.6.
Sawyer, Hall, Matthew J. Kauffman, Arthur D. Middleton, Thomas A. Morrison, Ryan M. Nielson, and Teal B. Wyckoff. 2013. A framework for understanding semi-permeable barrier effects on migratory ungulates.” Edited by Nathalie Pettorelli. Journal of Applied Ecology 50 (1): 68–78. https://doi.org/10.1111/1365-2664.12013.
Schroeder, Natalia M., Silvia D. Matteucci, Pablo G. Moreno, Pablo Gregorio, Ramiro Ovejero, Paula Taraborelli, and Pablo D. Carmanchahi. 2014. Spatial and Seasonal Dynamic of Abundance and Distribution of Guanaco and Livestock: Insights from Using Density Surface and Null Models.” PLOS ONE 9 (1): e85960. https://doi.org/10.1371/JOURNAL.PONE.0085960.
Scoones, Ian, Alec Bishi, Neo Mapitse, Rebone Moerane, Mary Louise Penrith, Ronny Sibanda, and Gavin Thomson. 2010. Foot-and-mouth disease and market access : challenges for the beef industry in southern Africa.” Pastoralism 1 (2): 135–64. https://doi.org/10.3362/2041-7136.2010.010.
Segar, Josiane, and Aidan Keane. 2020. Species and demographic responses to wildlife‐friendly fencing on ungulate crossing success and behavior.” Conservation Science and Practice 2 (10): e285. https://doi.org/10.1111/csp2.285.
Seigle‐Ferrand, J., P. Marchand, N. Morellet, J.‐M. Gaillard, A. J. M. Hewison, S. Saïd, Y. Chaval, et al. 2021. On this side of the fence: Functional responses to linear landscape features shape the home range of large herbivores.” Journal of Animal Ecology, November. https://doi.org/10.1111/1365-2656.13633.
Skinner, J. D., and Christian T. Chimimba. 2005. The Mammals of the Southern African Sub-region. Cambridge University Press. https://doi.org/10.1017/cbo9781107340992.
Spinage, C. A. 1992. The decline of the Kalahari wildebeest.” Oryx 26 (3): 147–50. https://doi.org/10.1017/S0030605300023577.
Sutmoller, Paul. 2002. The fencing issue relative to the control of foot-and-mouth disease.” In Annals of the New York Academy of Sciences, 969:191–200. 1. New York Academy of Sciences. https://doi.org/10.1111/j.1749-6632.2002.tb04377.x.
Taylor, W. Andrew, Peter A. Lindsey, Samantha K. Nicholson, Claire Relton, and Harriet T. Davies-Mostert. 2020. Jobs, game meat and profits: The benefits of wildlife ranching on marginal lands in South Africa.” Biological Conservation 245 (May): 108561. https://doi.org/10.1016/J.BIOCON.2020.108561.
Tinley, K. L. 1971. The Case For Saving Etosha.” African Wild Life 25 (1): 1–16.
Turpie, Authors Jane, Guy Midgley, Chris Brown, Jon Barnes, John Pallett, Phillip Desmet, Jacquie Tarr, and Peter Tarr. 2010. Climate Change Vulnerability and Adaptation Assessment for Namibia ’ s Biodiversity and Protected. Ministry of Environment and Tourism, Directorate of Parks & Wildlife Management. Ministry of Environment and Tourism, Directorate of Parks & Wildlife Management, Namibia Parks and Wildlife, no. May: 28.
Vanak, Abi Tamim, Maria Thaker, and Rob Slotow. 2010. Do fences create an edge-effect on the movement patterns of a highly mobile mega-herbivore? Biological Conservation 143 (11): 2631–37. https://doi.org/10.1016/j.biocon.2010.07.005.
Weaver, Genevieve V., Joseph Domenech, Alex R. Thiermann, and William B. Karesh. 2013. Foot and mouth disease: A look from the wild side.” Journal of Wildlife Diseases 49 (4): 759–85. https://doi.org/10.7589/2012-11-276.
Weise, Florian J., Quenton Wessels, Stuart Munro, and Matthew Solberg. 2011. Using Artificial Passageways to Facilitate the Movement of Wildlife on Namibian Farmland.” Https://Doi.org/10.3957/056.044.0213 44 (2): 161–66. https://doi.org/10.3957/056.044.0213.
Wilkinson, Christine E., Alex McInturff, Maggi Kelly, and Justin S. Brashares. 2021. Quantifying wildlife responses to conservation fencing in East Africa.” Biological Conservation 256 (April): 109071. https://doi.org/10.1016/j.biocon.2021.109071.
Xu, Wenjing, Nandintsetseg Dejid, Valentine Herrmann, Hall Sawyer, and Arthur D. Middleton. 2021. Barrier Behaviour Analysis (BaBA) reveals extensive effects of fencing on wide-ranging ungulates.” Edited by Matt Hayward. Journal of Applied Ecology, January, 1365–2664.13806. https://doi.org/10.1111/1365-2664.13806.
You, Zhang Qiang, Zhi Gang Jiang, Chun Wang Li, and David Mallon. 2013. Impacts of grassland fence on the behavior and habitat area of the critically endangered Przewalski’s gazelle around the Qinghai Lake.” Chinese Science Bulletin 58 (18): 2262–68. https://doi.org/10.1007/s11434-013-5844-9.